More Useless Geekbench Bullshit – NVIDIA “RTX 3080” Mobile Makes an Appearance – Surprise GA104 Scam – Now With 48SM’s and 16GB of GDDR6 Memory
6 January 2021
3070 3080 AMD ASUS GDDR6 GeForce jensen LAPTOP mobile NOTEBOOK Nvidia rtx Ryzen SCAM

I’ve said it before and I’ll say it again. Geekbench whether used to conduct testing on CPUs or GPUs is a garbage mobile benchmark best catered towards phone faggots. It’s CPU derived tests have always performed better on Intel processors (until it didn’t) as the powers of AVX merely couldn’t hold the substantial IPC gains that new Ryzen processors provide.

It’s garbage compute based graphics tests are just that, garbage irrelevant compute tests. But then again these kinds of fucking things make up half of NVIDIA’s marketing when comparing Ampere to previous generation Turing products so I guess that’s just fine and dandy.

Recently a leak was discovered on a Geekbench OpenCL compute result of the mobile variant of the NVIDIA GeForce RTX 3070 which came in underpowered being 27% slower than the desktop variant of the same graphics card due to seriously restrained clockspeeds and power limitations on the OpenCL compute scores alone while only managing to being 23% faster than a mobile RTX 2070 Super…. in a compute benchmark.

Actual practice in games would show an even worse ordeal for these low clocked mobile trash cards as they scale far more with clockspeed than mere compute synthetics.

However in the case of the mobile RTX 3070 it merely suffered a fate of a severely reduced clockspeed and possibly bandwidth as well. This is not the case for the mobile variant for the RTX 3080.

As yet another irrelevant Geekbench 5 OpenCL test has confirmed the existence of a mobile GeForce RTX 3080 and once again spotted by @TUM_APISAK the specifications of this mobile 3080 would probably kill you with laughter.

Just for context we previously saw the RTX 3070 mobile version offered us 40 SMs and 8GB of GDDR6 while the desktop variant offers 46 SMs indicating that NVIDIA are actually scraping off SM’s for their mobile variants and outright offering inferior products comparative to their desktop equivalents of that they are named.

The RTX 3080 on the desktop has 10GB of GDDR6X memory, along with 68 SM’s giving a total CUDA core count of 4352 or (8704) cores in total if you’re a retarded consoomer.

While the Geekbench 5 result shows us that the Mobile RTX 3080 offers just 48 SM’s which gives us 3072 CUDA cores or (6144) total cores while running at a maximum frequency of 1550MHz, uniquely paired with one of AMD’s latest Ryzen 5900H processors.

This is a core reduction of 29.4%. In no way shape or fucking form could one even dare call this an RTX 3080 because it isn’t. It shares more in fucking common with the 3070 than being a GA102 class chip which it probably is.

Because the severe lack of CUDA cores and SM’s isn’t the only major difference here is that the Geekbench page also shows that this mobile GPU has 16GB of VRAM. An astonishing amount and goes to show just how much Jensen Huang has fucked over original RTX Ampere buyers with insufficient amounts of onboard memory.

As I said before, this is most certainly a GA104 die rather than a GA102 which the real RTX 3080 is based upon. All the more likely that those 16 gigabytes of memory are probably standard GDDR6.

NVIDIA are literally packaging this fucking GA104 garbage and branding it as a 3080. I guess probably because Ampere is just too efficient to conceive the thought of packaging 628mm² of Samsung 8nm trash inside a notebook.

I really don’t know how to continue from here other than bitch about its performance results so let’s do that.

This fake RTX 3080 Mobile achieve an irrelevant score of 139,181 points in the GB5’s OpenCL compute. Compared to the mobile RTX 3070 that we saw just days ago this “3080” nets a 25.57% higher score despite only offering 8 SM’s more so than its little brother, though I’m sure the clockspeed advantages had a major role to play in the final scoring as well.

By comparison a desktop GeForce RTX 2080 Ti can potentially score anywhere from 140,000 to as high as 165,000 so on this compute oriented synthetic this supposed RTX “3080” for the most part is anywhere from 9.6% behind or up to 17.8% behind that of a desktop 2080Ti. Of course please do bare in mind that in actual gaming workloads (who even does that?!) that once again this mobile card is going to suffer in comparison.

A desktop 3070 offering 46 SM’s with a higher frequency topples out around 150,000 or thereabouts give or take 5-percent or so. With a score selected of 147,382 at a maximum frequency of 1.81GHz the desktop 3070 proves to be 5.89% faster in this compute oriented workload.

Entirely for shits and giggles in comparison with an AMD RX 6800 of which AMD has never had any real footing on these kinds of tests I come across four individual OpenCL scores that top out at 137,275.

So congratulations, in a synthetical compute benchmark your Mobile “3080” is 1.38% faster than a desktop RX 6800 which RDNA struggles with in comparison, at least your overpriced craptop comes with a free TV!

I’m glad to see mobile devices offering NVIDIA GeForce with suitable and far superior AMD Ryzen processors especially in “high-end” configurations such as this however I’m seriously disgusted by whilst very amused at this horrible attempt at what’s to be called an RTX 3080.

blog comments powered by Disqus